{
"$type": "site.standard.document",
"bskyPostRef": {
"cid": "bafyreie25jnju23vyrirghysju2zgaa6futpwajpbbcqqkyvcpdurwvnu4",
"uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mjgotzii24e2"
},
"path": "/t/bootstrapping-xml-schema-definitions-with-claude-opus-4-6-a-case-study-the-good-the-bad-and-the-ugly/13853#post_11",
"publishedAt": "2026-04-13T20:18:27.000Z",
"site": "https://discourse.haskell.org",
"textContent": "Great post - very relatable.\n\nLLMs aren’t failing loudly, they’re failing _convincingly_. The constant generator, bypassing self-hosting, even deleting tests - all classic “looks correct” over “is correct.”\n\nYour TL;DR is spot on: without a strong test harness, vibe coding breaks.\n\nFeels like the real skill now is designing systems where the AI _can’t cheat_.",
"title": "Bootstrapping XML schema definitions with Claude Opus 4.6: A case study (the good, the bad, and the ugly)"
}